ARL outlines its plans for HPC and data analytics
The lab released its S&T strategy for the next five years, which includes development in tactical high performance computing and large scale data analytics.
The Army Research Lab is planning to apply supercomputing muscle and large-scale data analytics in the process of supporting the Army’s mission in an increasingly complex environment, according to ARL’s recently released Science and Technology Strategy for 2015-2019.
The strategy, which follows an initial, overarching document released last year, outlines ARL’s plans in 24 program areas in all, and is aligned with Army Training and Doctrine Command's guidance, ARL said. Two of those areas—high-performance computing and data analysis—reflect the services efforts to, among other things, operate unmanned and autonomous systems and improve situational awareness through real-time analytics.
ARL outlined a three-pronged mission for an overarching computational sciences campaign that includes harnessing the potential of computational sciences and emerging high-performance computers to support systems incorporating predictive modeling and simulation modalities, facilitating information dominance, distributed maneuver operations, and conduct human sciences through computational data intensive sciences.
ARL envisions using HPC, at speeds up to up to 100 petaflops (100 quadrillion floating point operations per second), for operations such as autonomous systems and real-time data analytics for soldiers and intelligence analysts.
The tactical HPC strategy aims to integrate four particular research areas, the first of which is advanced computer research. Advanced computer research will facilitate the use of emerging architectures such as new algorithm designs and new analysis approaches that must be developed in order for the increase in computing capacity to be effective. Second, research into provisioning aforementioned systems under a distributed computing architecture will examine how to limit network hop through novel concepts. Third, dynamic binary translation research will aim to limit software re-writes for maximum performance and optimization in a runtime environment. And fourth, research into power and architecture aware computing for enhanced intelligence of provisioning systems will look into designing systems with greater awareness of computing capacity along with mission appropriateness.
The Army has identified numerous warfighting challenges that will match up to ARL’s HPC and large-scale data analytic remedies, such as development and sustainment of a high degree of situational understanding, assure uninterrupted access to critical communications and information links, conduct effective air-ground combined arms reconnaissance, conduct combined arms air-ground maneuvers, coordinate and integrate Army and joint interorganizational and multinational fires, conduct targeting across all domains and, finally, understand, visualize, describe, direct, lead, and assess operations.
In order to address these needs and goals outlined for a tactical HPC agenda, ARL identified necessary personnel needed to achieve their desired outcomes. Such personnel include mathematics experts for employing heuristic methodology to solve various paradigms, computer scientists and engineers as well as individuals to examine test cases that will determine methods to overcome performance inhibitors such as memory access patterns. In addition to personnel, ARL outlined the necessary infrastructure needed such as access to processor designs and HPC platforms for examining algorithm designs, code mapping and benchmark suite analysis.
ARL has high hopes in the realm of large-scale data analysis, which it plans to apply to vastly improved situational awareness, predictive analytics for decision making, enhanced autonomous technology, accelerated solider training utilizing live and virtual data analytics, and spurring new innovations for Army systems with observational, experimental and simulations data.
ARL wants to develop a scalable computational method through large parallel hierarchical computing architectures in order to advance inherent large-scale complex data.
Similar to the HPC framework, ARL outlined personnel necessary large scale data analysis, including computational mathematics experts, computational informatics experts and computer science and engineering experts. These skilled tacticians will work with the same infrastructure specifics as personnel in tactical HPC.
Identified goals in specific areas of large-scale data analytics include scalable mathematical algorithms, data-enabled science, predictive computational methods, real-time data analytics, model order reduction, human cognition based mathematical approaches, neuro and biologically inspired methods, science analyzing large-scale data from wearable electronics/technologies and sensing, along with human and artificial intelligence.